Amazon working on ‘rigged’ voices for Alexa

Alexa, Amazon’s smart assistant, could receive a feature that allows it to mimic a person’s voice based on one minute of audio.

Amazon says it is preparing a feature that allows Alexa to mimic a human voice from one minute of audio. The company announced it at its re:MARS conference held in Las Vegas in the United States. The technology in question is presented as an opportunity to remember loved ones who have passed away by using their voice to read texts.

Regardless of the psychological consequences of this idea and the ethical questions that arise about the rights to use someone’s voice, even a deceased person’s, the technology, which is capable of quickly and easily falsifying someone’s voice, obviously has other possible effects. Therefore, it could also be used for fraudulent purposes, such as voice phishing (“vishing”). Think, for example, of a call from the CFO of a company asking an employee to immediately deposit an amount into a specific account (belonging to scammers).

For the sake of clarity, this is a demo of possible future technology. Computer systems can now mimic voices, but these days they still require much more information than a minute of sound.

Amazon showed off the feature in a video, where a child asks his grandmother to ‘read’ him a story. It is not known what stage of development the feature is in. The demo appears to be based on advances in Amazon’s text-to-speech technology, as explained in this White book.

Amazon says it is preparing a feature that allows Alexa to mimic a human voice from one minute of audio. The company announced it at its re:MARS conference held in Las Vegas in the United States. The technology in question is presented as an opportunity to remember deceased loved ones by using their voices to read texts, regardless of the psychological consequences of this idea and the ethical questions that arise about the rights to use someone’s voice. Quickly and easily faking someone’s voice obviously has other possible effects. Therefore, it could also be used for fraudulent purposes, such as voice phishing (“vishing”). Think, for example, of a call from the CFO of a company asking an employee to deposit an amount without delay into a specific account (belonging to scammers). For the sake of clarity, this is a demonstration of a possible technology future. The computer systems that are currently present imitate voices, but they currently need more information than just a minute. grandmother. It is not known what stage of development the feature is in. The demo appears to be based on advances in Amazon’s text-to-speech technology, as explained in this white paper.

Leave a Comment